Foundation: A broad review of advances in PD research and technologies (2013 - early 2025) was conducted, with a focus on risk assessment, disease progression, medication response, and digital approaches to symptom tracking, including wearables. Datasets of value, including the Parkinson's Progression Markers Initiative (PPMI), the Critical Path for Parkinson’s (CPP), and the UK Biobank, were accessed, while a data harmonisation strategy based on the OMOP common data model was set up and applied on third-party (PPMI) and project-collected clinical data (dBM-DEV study). To steer R&D activities, a Trustworthy AI development framework was created and kept up to date, accounting for relevant regulations and guidelines.
User research and co-creation: Towards efficient stakeholder engagement, a patient panel was formed and user research was conducted with the panel and HCPs. Two web surveys further expanded stakeholder feedback. These activities helped in identifying key needs and attitudes toward the project’s digital health tools, the mAI-Health and mAI-Care patient-facing apps for PD screening and management, respectively, and the mAI-Insights platform for HCPs. Co-creation activities with stakeholders, including workshops on trustworthy AI qualities and user experience co-design sessions, shaped the first minimum viable product version (MVP) of the tools. Over 500 persons were involved in participatory activities.
Data analysis and AI: Digital biomarkers were developed to passively track REM sleep behaviour disorder (RBD), daytime sleepiness, rest tremor, slowness of movement, dyskinesia, and physical activity using multiple data streams, including wrist wearables and smartphone typing patterns. Active tests were also created, comprising mobile versions of standardised cognitive tasks and a motor function test using smartphone camera-based pose tracking to assess posture, balance, and walking. Development relied on third-party datasets and the project’s dBM-DEV study data. For PD risk assessment, a model was developed to detect PD based on smartwatch data-derived sleep and activity metrics of the Verily Study Watch cohort, combined with established criteria for prodromal PD. For disease prognosis, predictive models for short-term (1 year) and long-term PD progression were developed using multimodal data from the PPMI and the PD Biomarkers Programme. Models predicting the onset of medication side effects, such as dyskinesias, were also created using the CPP cohorts’ clinico-demographic and medication data. Genetic variations linked to PD risk and progression were further explored in the AMP-PD cohorts to inform the predictive models.
Digital health ecosystem: Through mapping user needs to system features, refinement based on co-creation feedback, and integration of validated research outcomes, mature versions of the AI-PROGNOSIS digital health tools were developed. The mAI-Health and mAI-Care apps feature a core smartphone and smartwatch data collection layer and incorporate tailored patient-reported outcomes, physician connectivity, and PD educational content. mAI-Insights includes patient connectivity, medical data logging, and digital biomarker-based remote symptom monitoring, while PD risk and progression tracking, as well as medication decision support, are enabled by the AI-PROGNOSIS predictive models. Α secure, orchestrated Cloud data storage and processing system serves as the backbone.
Clinical studies: The dBM-DEV study (NCT06444789, 80 participants) was completed and the AI-PMP study was initiated (NCT07189468). dBM-DEV collected clinico-demographic, smartwatch, typing, and motor/cognitive function digital test data, alongside sleep lab recordings from persons with RBD, persons with PD (PwPD) and healthy individuals, to extend and validate the digital biomarkers of the project. AI-PMP will follow 100 PwPD over a year, collecting similar data enriched with genetic information, aiming primarily to validate the project’s PD progression models. The protocol of a third project study, AI-PRA, on the validation of the PD risk assessment system was established and is expected to initiate in early 2026.
